A “Zero Noise” photon counting algorithm for enhancing ICMOS detection sensitivity under extremely low-light conditions

Abstract

ICMOS detectors typically rely on analog integration methods for signal processing in low-light imaging scenarios. However, under extremely low-light conditions at the single-photon level, their performance is still constrained by low photoelectric conversion efficiency and interference from system noise. Existing studies often adopt the centroid method to achieve photon counting imaging, which can effectively suppress readout noise from the system backend but offers limited suppression of non-Gaussian noise at the frontend, such as dark noise and shot noise, resulting in a clear bottleneck in overall detection performance. To address this issue, this paper proposes a “Zero Noise” photon counting algorithm specifically designed to effectively suppress non-Gaussian noise from the system frontend. The method first reduces readout noise through Gaussian fitting and an 8-connected seed-filling algorithm, then constructs a photon confidence interval $R$ by combining photon counting statistical modeling with the Kolmogorov–Smirnov (KS) test, which is used for noise suppression and image reconstruction. To verify the effectiveness of the proposed approach, comparative experiments were conducted under two scenarios: active laser detection and passive imaging of a target board, using the traditional analog integration method and the centroid method as baselines. The transverse photon counting (TPC) statistical curve was used to calculate image contrast and evaluate the improvement in Signal-to-Noise Ratio (SNR). Experimental results show that, compared with existing traditional methods, the proposed algorithm significantly improves detection sensitivity under extremely low-light conditions and demonstrates superior overall performance.